AI DUNN Right Weekly - Issue #8
- Nov 18, 2025
- 10 min read
Practical AI insights for business growth
Hey AI Innovators! 👋
Welcome back to AI DUNN Right Weekly! This week Chinese hackers weaponized AI for mass cyber espionage, Cursor raised $2.3 billion on questionable unit economics, and LinkedIn finally made finding people actually work.
Here’s what matters for your business:
• Chinese state hackers used Claude Code to infiltrate 30+ organizations with 80-90% AI automation
• Cursor’s $30 billion valuation reveals the AI middleware profit margin problem
• LinkedIn launches AI-powered search that understands conversational queries
• ChatGPT copywriting framework generates high-conversion ads using 9 styles and 5 tones
• AI-driven heatmap analysis optimizes landing pages before spending test budgets
Read time: 5 minutes
🚀 This Week’s Game Changer
AI Becomes the Weapon - Chinese Hackers Automate Cyber Espionage at Scale
What happened: Anthropic disclosed that Chinese state-sponsored hackers used Claude Code to infiltrate 30+ organizations including banks, tech firms, chemical manufacturers, and government bodies. The AI handled 80-90% of the work autonomously with minimal human oversight, running through four distinct attack phases: target selection, vulnerability scanning, iterative exploitation, and credential theft.
The innovation: This ain’t just another hacking story - it’s proof that AI-driven espionage works at scale. The hackers essentially created an autonomous security penetration system that could assess, adapt, and exploit vulnerabilities faster than traditional methods. Anthropic caught it through usage pattern analysis and published a detailed flowchart showing exactly how the attack sequence worked (which is either transparency or accidentally providing an instruction manual, depending on your perspective).
Business impact:
For All Businesses: The gap between AI-enabled attackers and traditional defense widens daily - if you’re not using AI for security, you’re behind
For Corporate Teams: Cybersecurity basics (strong passwords, two-factor authentication) matter more than ever when AI can automate vulnerability discovery
For Tech Leaders: Budget constraints on security talent create exploitable weaknesses - UK civil service caps cybersecurity salaries at £100k while American talent commands £2M+
The bigger picture: Kyle’s take resonates: “This signals AI-driven espionage is feasible at scale, but let’s be honest, it’s probably been happening for ages and we just found out.” Anthropic’s disclosure walks a weird line between damage control and inadvertent marketing: “Our tools are so good, Chinese state-backed hackers choose them!” It’s technically impressive while being completely the wrong endorsement.
Why it matters: The AI wars started while everyone focused on productivity gains. When automation can spend days iteratively probing systems without human fatigue or detection patterns, traditional security assumptions break. The lesson for every business leader: defensive AI needs to match offensive AI, or you’re just hoping attackers choose easier targets. Hope ain’t strategy.
🛠 AI Tool Spotlight
LinkedIn AI Search - Natural Language Discovery That Actually Works
What it does: LinkedIn launched AI-powered people search for Premium members that finally eliminates their historically terrible filter system. Instead of clicking through endless dropdown menus hoping to guess the exact job title someone uses, you simply ask in plain English. Want “healthcare investors with FDA background”? Just type it. Looking for “NYC productivity startup co-founders”? Done. The system understands intent rather than demanding precise keyword matches.
Key features:
Conversational queries replace rigid keyword matching
Multi-factor searches work in one simple box instead of toggling between company, location, industry filters
Invisible talent surfaces through contextual understanding - finds advisors, founders, specialists missed by title-based searches
“I’m looking for…” prompt appears in search bar making intent-based discovery intuitive
Premium-only during initial U.S. rollout with global expansion coming within months
Results quality varies - broad queries like “voice AI founders” show random matches while specific asks like “Y Combinator alumni” hit targets
Real example: Traditional LinkedIn search required knowing exact job titles, current companies, and specific skills tags. Someone working as “Head of Growth” at one company might list “VP Marketing” at another. You’d miss them entirely if your filter said “Chief Marketing Officer.” AI search understands these are related roles and surfaces all relevant profiles when you ask for “senior marketing leaders in SaaS.”
The workflow:
Click the search bar (Premium members see “I’m looking for…” prompt)
Describe who you need in conversational language
Review results ranked by relevance to your intent
Refine query using natural follow-ups if needed
Connect with discovered professionals
Best use cases: Finding investment partners who understand your specific industry challenges, recruiting for roles where job titles vary wildly between companies, discovering advisors with niche expertise combinations that don’t match standard categories, sourcing partnerships with complementary capabilities. Particularly valuable for networkers who know what expertise they need but don’t know what it’s called in LinkedIn’s taxonomy.
Why it’s a breakthrough: LinkedIn’s massive professional database fought against itself for years because terrible search required exact phrases nobody actually knows. You’d give up after three failed filter combinations or settle for barely-relevant connections. Plain language eliminates that friction entirely. Users find connections speaking normally versus learning database query syntax. Professional discovery just became accessible to people who need relationships but lack LinkedIn search PhD credentials.
⚡ The 5-Minute AI Academy
The AI Middleware Profit Margin Problem - Cursor’s $30 Billion Question
The difference between valuable AI companies and expensive middlemen often comes down to unit economics. Cursor’s $2.3 billion raise at $30 billion valuation looks impressive until you examine where revenue actually goes. Most AI tool businesses face this same challenge - are they creating sustainable value or just making someone else’s AI prettier?
The Problem: Cursor generates $1 billion in revenue, which sounds fantastic. They’re the leading AI coding IDE, producing “more code than any other agent in the world” according to their metrics. But here’s the catch: most of that revenue passes straight through to model providers. OpenAI, Anthropic, and Google supply the underlying intelligence while Cursor provides the interface. Reports suggest margin on model usage hits only 20%, meaning for every dollar Cursor earns, they’re spending 80 cents on AI API calls.
The Competition: ByteDance just launched an AI coding agent for $1.30/month (10 yuan) that includes Claude Code Pro access. Sure, it’s Chinese-interface only and targeting local markets, but it reveals how quickly pricing collapses when the underlying technology is commoditized. When your core value is “we made GPT easier to use,” you’re vulnerable to the AI company adding those features directly or competitors undercutting your pricing by 95%.
The 5 Strategic Questions:
1. What Do You Actually Own?
Middleware businesses often don’t own the intelligence, the data, or the unique workflow. Cursor built a fantastic developer experience, but that’s defensible only if switching costs exceed competitor advantages. Kyle’s observation hits hard: “They’re basically a shopfront charging us then paying most of it to Claude and GPT. They’re a middleman with a nice interface.”
2. Where’s Your Pricing Power?
When ByteDance offers similar functionality at $1.30 while you charge $20-30 monthly, your pricing reflects brand and current market position, not cost structure or competitive moat. The moment customers perceive alternatives as equivalent, price becomes the deciding factor. Middleware rarely controls pricing leverage long-term.
3. What Prevents Disintermediation?
OpenAI and Anthropic see exactly how much money flows through tools like Cursor. They understand the revenue potential. What stops them from launching competitive features and capturing margin currently going to middleware? Usually nothing beyond focus and development priorities. When those priorities shift, middleware businesses face existential threats overnight.
4. Can You Create Switching Costs?
Real defensibility comes from accumulated user investment that makes switching painful. GitHub integration, custom workflows, team collaboration patterns, saved configurations. These create stickiness beyond features. The question: are your switching costs stronger than competitors’ price advantages? Cursor’s developer-friendly interface builds some moat, but is it $30 billion worth?
5. Where’s Your Data Moat?
The strongest AI businesses own proprietary datasets that improve model performance in ways competitors can’t replicate. Cursor sees how developers use AI for coding. That usage data could train better models, but they don’t own the underlying intelligence. They’re feeding OpenAI and Anthropic the data that makes competing models better, which… doesn’t sound like a moat.
The Lesson for Your Business:
If you’re building anything AI-related, these questions matter immediately. Don’t just be a prettier interface for ChatGPT. Don’t rely on making Claude easier to use.
Focus on:
Owning unique data that improves over time with usage
Controlling relationships that create network effects or community moats
Building workflow integration so deep that switching means rebuilding processes
Creating vertical specialization where generic AI tools don’t match domain-specific optimization
Action step: If you’re currently building or planning an AI business, honestly answer those five strategic questions. If your answers feel uncomfortable, that’s valuable information now rather than painful discovery later. The best time to adjust positioning is before raising capital at valuations that require defending questionable unit economics.
Expected reality: Many AI middleware businesses will consolidate or disappear as model providers expand features and pricing pressure intensifies. The survivors will own something beyond interface design - whether that’s data, relationships, workflows, or domain expertise that creates genuine defensibility.
💡 Prompt of the Week
AI Landing Page Heatmap Analysis
Stop guessing which landing page elements convert and which waste attention. This prompt creates systematic optimization frameworks using AI-powered heatmap analysis:
Why this works: Most businesses waste testing budgets running random A/B experiments hoping something improves conversion. Heatmap analysis shows exactly where attention goes, which elements cause confusion, and where visitors abandon. AI amplifies this by identifying patterns across thousands of sessions faster than human analysis. You move from guessing to targeting specific friction points backed by behavioral data.
Best use case: E-commerce product pages, SaaS signup flows, webinar registration pages, lead capture funnels - anywhere conversion rate directly impacts revenue and you have enough traffic volume to generate meaningful heatmap data (minimum 1000 monthly visitors recommended).
🧠 Quick Wins: 5 AI Tools Worth Investigating
Based on this week’s newsletter coverage and emerging capabilities:
🔍 LinkedIn AI Search - Conversational people discovery for Premium membersUse case: Finding investors with niche expertise combinations, recruiting when job titles vary, sourcing partnerships, discovering advisors with specific background requirements
🎨 ChatGPT Universal Copy Prompt - 9 writing styles and 5 tones in systematic frameworkUse case: Creating ad copy, email sequences, landing pages, content that converts without hiring copywriters, testing message variations at scale
🎵 Udio AI Music - Complete song generation with vocals and productionUse case: Custom podcast theme music, branded audio content, video soundtracks, presentation background music without licensing fees
📊 Kimi K2 - 1M+ token context window for document analysisUse case: Analyzing entire codebases, comparing lengthy contracts, reviewing research papers, multilingual document understanding without context loss
⚙️ Runable No-Code Platform - Visual AI workflow builder connecting multiple modelsUse case: Building customer support systems, lead qualification automation, proposal generators, data enrichment pipelines without engineering teams
📈 Business Intel: This Week’s Market Movers
🔐 AI Weaponization Goes Mainstream Chinese hackers using Claude Code for mass espionage proved AI-driven attacks work at scale. Anthropic’s disclosure showed 80-90% autonomous operation infiltrating 30+ organizations. The strategic implication: defensive capabilities must match offensive AI sophistication or security becomes hoping you’re not the easiest target.
💰 Cursor Raises $2.3B Despite Margin Questions AI coding IDE Cursor hit $30 billion valuation on $1 billion revenue, attracting investment from Accel, Andreessen Horowitz, Nvidia, and Google. The unit economics raise concerns: 80% of revenue reportedly passes to model providers (OpenAI, Anthropic, Google), leaving 20% margins that compress when competitors undercut pricing. ByteDance’s $1.30/month alternative highlights pricing vulnerability.
🔍 LinkedIn Eliminates Search Friction Premium members gain AI-powered people search that understands conversational queries instead of requiring exact keyword matches. The accessibility shift opens previously invisible professional networks to users who understand needs but lack LinkedIn’s search syntax expertise. Smart search becomes table stakes rather than premium differentiator once adoption proves user retention value.
📝 Universal Copywriting Framework Launched ChatGPT Central released systematic copywriting prompt supporting 9 styles and 5 tones with structured frameworks. MIT research showed AI-assisted writing produces faster completion and higher quality drafts, with largest gains for less experienced writers. McKinsey’s 2024 survey found 65% of firms using generative AI, with written content as top use case.
📉 Free Tools Continue Disrupting Premium Markets Pattern accelerates where AI automation subsidizes zero-cost user acquisition. When technology eliminates 70% of operational costs through automated support and quality assurance, giving away core products becomes profitable through upsell to high-margin AI features. Canva’s Affinity Suite giveaway captured 1M+ users in 48 hours following this playbook.
📚 This Week’s Curated Reading
Based on key developments from this week’s AI news:
• The Middleware Margin Problem: Cursor’s valuation highlights the challenge facing AI tool businesses that don’t own underlying intelligence. When your core value is interface design and 80% of revenue passes to model providers, you’re vulnerable to disintermediation and pricing pressure from both suppliers and competitors.
• Security Asymmetry Widens: AI-powered offensive capabilities outpace defensive adoption. When attackers automate vulnerability discovery and exploitation while defenders rely on traditional methods, the capability gap creates systematic risk that budget constraints and talent shortages amplify.
• Accessibility Eliminates Moats: LinkedIn’s AI search demonstrates how conversational interfaces collapse expertise barriers that previously created defensibility. When you no longer need to learn complicated syntax to access value, user bases expand beyond expert users to mainstream adoption.
• AI SEO Strategy Shift: Human-generated source material amplified through AI beats mass-produced AI content. Google’s algorithm increasingly detects and penalizes pure AI spam. The winning approach: create genuine expertise through conversation (transcripts, videos, discussions), then use AI to expand and optimize that authentic content.
• Unit Economics Trump Features: Free isn’t sustainable without AI-powered cost transformation. Businesses successfully offering zero-cost tiers automate 70%+ of support, onboarding, and quality assurance. The moat isn’t software features - it’s operational efficiency that makes free profitable through high-margin upsell.
🎯 Action Items for This Week
For Corporate Teams:
Audit your AI tool stack for middleware businesses with questionable margins - are you paying premium prices for features model providers will likely offer directly?
Review cybersecurity basics (password policies, two-factor authentication) knowing AI-powered attacks now automate what previously required human expertise
Test LinkedIn AI search if Premium subscriptions exist - identify whether natural language discovery improves networking effectiveness
For Small Businesses:
Identify one repetitive copywriting task (emails, social posts, ad variations) and implement the ChatGPT universal copy framework
Calculate whether AI automation could transform operational costs enough to support free tier acquisition strategy
Examine landing pages using free heatmap tools (Microsoft Clarity) to identify conversion friction points before testing
For Entrepreneurs:
If building AI businesses, honestly answer the five strategic questions about defensibility before committing to middleware positioning
Experiment with AI tools solving specific problems rather than general assistants - specialization creates stronger moats
Consider how AI SEO strategy (human source content + AI amplification) applies to your content marketing approach
🔮 Looking Ahead
Next week’s AI DUNN Right Weekly will cover:
Voice AI agents moving from pilot to production - what the 2025 State of Voice AI Report reveals
AI-driven browser context visualization for real-time content adaptation
Building sustainable AI businesses beyond interface middleware
How professionals are measuring actual ROI on AI tool investments rather than trusting hype
Have questions or topic requests? Reply to this email. I read every message and use your feedback to shape future issues.
That’s a wrap for Issue #8!
This week proved AI crossed from productivity tool to strategic weapon, middleware businesses face margin pressure they can’t easily solve, and accessibility features that eliminate expertise barriers can destroy decade-old moats overnight. The businesses that understand unit economics, own defensible assets beyond interface design, and implement AI faster than competitors attack them will compound advantages. Everyone else hopes being fast follower stays viable long enough.
Stay innovative,Jackie @ AI DUNN Right
P.S. - That middleware margin problem? If you’re paying for AI tools, check whether you’re essentially funding a nice UI for ChatGPT or Claude. Not saying pretty interfaces lack value, just suggesting you calculate whether you’re paying $20/month for what could be $2/month of API calls plus some CSS. Sometimes the prettiest solution ain’t the smartest purchase.
Comments